MBE growth of 2H-MoTe2 and 1T'-MoTe2 on 3D substrates

TL;DR

This study explores the molecular beam epitaxy growth of 2H and 1T' phases of MoTe2 on 3D substrates, analyzing phase boundaries, grain size, and stability to advance understanding of this less-explored material.

Contribution

It demonstrates controlled MBE growth of MoTe2 phases, characterizes their structural properties, and investigates their stability and phase boundaries on different substrates.

Findings

Successful MBE growth of 2H and 1T' MoTe2 phases.

Identification of grain size and twinned grains via GI-XRD.

Observation of excess Te incorporation affecting electrical properties.

Abstract

MoTe2 is the least explored material in the Molybdenum-chalcogen family, which crystallizes in thermodynamically stable semiconducting 2H phase at \textless 500 C and 1T' metallic phase at higher temperatures. Molecular beam epitaxy (MBE) provides an unique opportunity to tackle the small electronegativity difference between Mo and Te while growing layer by layer away from thermodynamic equilibrium. For a few-layer MoTe2 grown at a moderate rate of $\sim$6 mins per monolayer under varied Te:Mo flux ratio and substrate temperature, the boundary between the 2 phases in MBE grown MoTe2 on CaF2 is characterized using Reflection high-energy electron diffraction (RHEED), Raman spectroscopy and X-ray photoemission spectroscopy (XPS). Grazing incidence X-ray diffraction (GI-XRD) reveals a grain size of $\sim$90 {\AA} and presence of twinned grains. XRD, transmission electron miscroscopy, RHEED, low energy electron diffraction along with lack of electrical conductivity modulation by field effect in MBE 2H-MoTe2 on GaAs (111) B show likelihood of excess Te incorporation in the films. Finally, thermal stability and air sensitivity of MBE 2H-MoTe2 is investigated by temperature dependent XRD and XPS, respectively.

Figures

a shows the crystal structure of a monolayer transition metal dichalcogenide with undistorted octahedral transition metal coordination. Lines of metal atoms run in three different directions (referred to as 1, 2, and 3; see inset of Fig. 8b), where each line is separated by a distance d. In the distorted structure, shown in Fig. 8b, the metal atoms shift, creating zigzag chains of metal atoms where two neighboring lines are separated by a smaller distance d-$\delta$. These chains run along a single direction (in this case, direction 1), causing the unit cell to change from hexagonal to rectang

Tables

Table 1. Table 1: Growth conditions for all samples in this study

Series	Sample ID	Substrate	MoTe2 Phase	Substrate Temperature ($°$C)	Te Flux (Torr)	Te:Mo flux ratio	Growth duration (mins)	Post growth anneal
A	A	CaF2 (111)	2H	340	6.5x10-6	297	30	none
	B	CaF2 (111)	2H+1T’	340	2.0x10-6	98	30	none
	C	CaF2 (111)	1T’	400	1.4x10-6	71	30	none
B	D	GaAs (111) B	2H	340	6.9x10-6	175	30	at 380$°$C for 10 mins without Te
	E	GaAs (111) B	2H	340	6.9x10-6	262	20	at 450$°$C for 3 mins and 550$°$C for 7 mins under Te

Table 2. Table 2: Positions of the various Raman peaks compared to measured values for bulk 2 H -MoTe 2 and reported values for 1 T’ -MoTe 2 Park  et al. ( 2015 ) .

Sample ID	A1g position (cm-1)	E22g position (cm-1)	Bg position (cm-1)	Ag position (cm-1)
Bulk 2H	174.47	235.80	NA	NA
A	173.30	235.79	NA	NA
B	174.14	236.66	158.07	255.59
C	NA	NA	159.14	256.23
1T’Park et al. (2015)	NA	NA	163	256.1

Table 3. Table 3: XPS peak positions for Mo 3d 5/2 and Te 3d 5/2 as well as Te:Mo ratio corresponding to both phases 2 H -MoTe 2 and 1 T’ -MoTe 2 compared to reported values

Sample ID	Te:Mo flux ratio	Mo 3d5/2 2H (eV)	Te 3d5/2 2H (eV)	Mo 3d5/2 1T’ (eV)	Te 3d5/2 1T’ (eV)	Te:Mo2H ratio w/o corr.	Te:Mo2H ratio corr.	Te:Mo1T’ ratio w/o corr.	Te:Mo1T’ ratio corr.
Bulk 2H		227.8 (air exposedJaegermann and Schmeisser (1986)) 228.0 (as cleavedJaegermann and Schmeisser (1986)) 227.8 (as cleavedBernede et al. (2003))	572.9 (air exposedJaegermann and Schmeisser (1986)) 573.0 (as cleavedJaegermann and Schmeisser (1986)) 572.4 (as cleavedBernede et al. (2003))				2.03
A	297	227.9	572.8				2.57
B	98	227.9	572.9	227.4	572.6	2.03	2.54	2.12	2.66
C	71	228	572.9	227.5	572.6	2.04	2.61	2.13	2.72
1T’				227.7Naylor et al. (2016) 228Zhou et al. (2015)	572.1Naylor et al. (2016) 572.6Zhou et al. (2015)

Table 4. Table 4: In-plane and out of plane lattice constants for the 2 H -MoTe 2 samples in this study obtained by various techniques

Sample ID	RHEED (Å)	XRD (Å)	GI-XRD (Å)	LEED (Å)	TEM (Å)
A	a = 3.5 $\pm$ 0.1		a = 3.64 $\pm$ 0.03 c = 14.4 $\pm$ 0.03		c = 13.9 $\pm$ 0.1
D	a = 3.6 $\pm$ 0.05	c = 14.52 $\pm$ 0.05			c = 13.8 $\pm$ 0.1
E	a = 3.5 $\pm$ 0.1	c = 14.25 $\pm$ 0.04		a = 3.57 $\pm$ 0.03
Bulk 2H		a = 3.52085Keum et al. (2015) c = 13.9664Keum et al. (2015)
Bulk 1T’		a = 6.3274Keum et al. (2015) b = 3.4755Keum et al. (2015) c = 13.8100Keum et al. (2015)

Full text

Transition metal chalcogenide, MoSe2, MoTe2

I Introduction

MoTe2 is still a relatively unexplored transitional metal dichalcogenide (TMD) and holds great promise. MoTe2 exists in trigonal prismatic (2H structure) semiconducting phase at room temperature and metallic 1T’ phase (monoclinic structure) at high temperatures or metallic Td phase (orthorhombic structure) when metastable 1T’ is cooled to $-$33$\degree$C-$-$13$\degree$CDeng et al. (2016). 1T’ and Td crystal structures have the same in-plane crystal structures but vary in vertical stacking. Monolayer 2H-MoTe2 is predicted to be among the smallest bandgap semiconducting TMDsGong et al. (2013). Using scanning tunneling spectroscopy (STS) measurements, the bandgap of monolayer 2H-MoTe2 has been measured to be between 1.03 eVDiaz et al. (2016) to 1.4 eVChen et al. (2016). 2H-MoTe2 has a close to broken band alignment with materials like SnSe2Schlaf et al. (1999), making them attractive for tunnel based devices such as Esaki diodesYan et al. (2015) and two-dimensional heterojunction interlayer tunneling field effect transistors (Thin-TFETs)Li et al. (2015, 2014). The 1T’ structure is metastable at room temperature when the bulk MoTe2 crystal is quenched rapidly. The Td-MoTe2 is a type II Weyl semimetalDeng et al. (2016), which is a new class of topological material. Traditionally, the transition between 2H and 1T’ phase was thought to be abrupt with respect to temperatureBrewer and Lamoreaux (1990) but recently a mixed phase region has been observed in the phase diagramKeum et al. (2015). The transition between the 2H and 1T’ phase holds promise for applications such as low resistance contactsCho et al. (2015) and phase change memory.

Large area, phase-pure epitaxial growth with layer controllability would enable industrial applications, but growth of MoTe2 is especially challenging, since the electronegativity difference between Mo and Te is just 0.3 eVBernede et al. (2003); Jaegermann and Schmeisser (1986); Gamble (1974), resulting in a weak bond. Until recently, MoTe2 has been obtained using tellurization of e-beam deposited molybdenum or molybdenum oxide filmsZhou et al. (2015); Park et al. (2015); Zhou et al. (2017) or by chemical vapor transport (CVT)Bernede et al. (1996); Keum et al. (2015). In case of tellurizing Mo, 1T’-MoTe2 phase is initially formed at 650$\degree$C, which can be converted to 2H-MoTe2 by a 3 hour anneal under Te vaporPark et al. (2015). Tellurizing MoO3 transforms to 2H-MoTe2 more readily, but one cannot be sure in achieving 100% percent reduction of oxygen. For CVT grown MoTe2 a mixed phase is observed in the growth temperature range of 500$\degree$C to 900$\degree$C based on the tellurium content in MoTexKeum et al. (2015), but under ultrahigh vacuum (UHV) in non-thermodynamic equilibrium conditions, this phase boundary is unknown. For example, the formation of a new metallic nanowire phase has been recently reported upon Te loss by annealing at 400-500C in UHVZhu et al. (2017). We recently demonstrated growth of 2H-MoTe2 using molecular beam epitaxy (MBE) in a superlattice with MoSe2 as well as Bi2Te3Vishwanath et al. (2016), where we used a growth temperature of 380$\degree$C. Growth temperature of $\sim$340$\degree$CRoy et al. (2016) or 200$\degree$CDiaz et al. (2016); Chen et al. (2016) have been used in other recent MBE demonstrations of 2H-MoTe2 growth.

In this study, we chose 340$\degree$C to be the lower bound of growth temperature, so as to keep the growth temperature higher than the Te cell temperature ($\sim$300$\degree$C). This avoids intentional accumulation of Te, while keeping the growth temperature significantly lower than the lower bound (500$\degree$C) of the mixed phase, as suggested by the phase diagram under 1 ATM of Te vaporKeum et al. (2015). We observe that even at a substrate growth temperature as low as 340$\degree$C, the crystalline phase of the MBE-grown MoTe2 has a sensitive dependence on Te flux in a Te rich environment. Therefore, in Section A of this paper, we present a series of 3 samples of MoTe2 under different growth conditions and analyze the effect of substrate temperature and tellurium flux on the MoTe2 crystal structure and stoichiometry. Here, we show that it is indeed feasible to grow phase-pure 2H-MoTe2 on a CaF2 substrate without any requirement of a post-growth anneal. In Section B we present the growth and electrical characterization of 2H-MoTe2 on GaAs (111)B. The transition to GaAs was motivated by the availability of high quality epi-ready n$+$ GaAs substrates necessary for characterization using techniques such as low energy electron diffraction (LEED). Table.1 summarizes the growth conditions of all the 5 samples in this study. The Mo flux is calculated using the experimentally determined growth rates while assuming zero desorption for molybdenum adatoms, which enables calculation of the Te:Mo flux ratio tabulated in Table.1 for all samples. Growth rate on CaF2 is calculated using cross-section transmission electron microscopy image and on GaAs (111) B from Reflection high-energy electron diffraction (RHEED) oscillations.

[FIGURE:]

II Section A : MoTe2 on CaF2

II.1 Growth conditions

Three samples (see Table.1) constitute the series of samples grown on CaF2. Calcium fluoride was chosen as a substrate because (i) it has an inert fluorine-terminated surface on which we have successfully grown MoSe2Vishwanath et al. (2015) and (ii) it provides a cavity effect enhancing the Raman signal due to its optical transparency. On the other hand, the Raman signal from MoTe2 grown on GaAs(111)B is very weakVishwanath et al. (2016). These samples were grown in a Riber 32 MBE system using elemental Mo delivered from an e-beam evaporator and elemental uncracked Te from a Knudsen cell. The growth duration was 30 mins for each sample and the Mo flux, which limits the growth rate, was set to $\sim$0.17 monolayer (ML) per minute or 6 minutes per ML. Supplementary (SI) Fig.1 shows that all CaF2 substrates were first heated to 800$\degree$C, held for 30 mins in order to anneal and degas. Sharp RHEED streaks of CaF2 prior to start of growth (see Fig.1(a)), show the smooth crystalline post-anneal growth surface. Then the substrates were lowered to respective growth temperatures, stabilized for $\sim$30 mins prior to thin film growth. The growth conditions are listed in Table.1. Using these 3 samples, we observe that, although the Te flux is 2 orders of magnitude higher than the Mo flux, the Te flux range to obtain 2H-MoTe2 is narrow and the substrate temperature control is critical. None of the MoTe2 films on CaF2 have been annealed in order to avoid phase change during annealing. All temperatures given in this study are thermocouple temperatures and the sample surface temperature is estimated about 20$\degree$C lower than the thermocouple temperature.

II.2 Results and discussion

II.2.1 Reflection high-energy electron diffraction (RHEED)

As seen in Fig.1(c), annealed CaF2 has a strong streaky RHEED pattern prior to growth. Sample A shows a more diffused but still streaky RHEED pattern, whereas sample B shows a mixed pattern comprising of streaks and a ring, which evolves into only a ring in sample C. Streaky RHEED points to layered growth with minimal mosaicity of the as-grown film, but progressive inclusion of a ring points to another growth mechanism taking over, which results in polycrystalline growth. Whether the polycrystalline material is the same phase as the streaky film or a different phase is elucidated through employing Raman and XPS characterization (described below). The RHEED streaks of MoTe2 in sample A along <11$\bar{2}$0> appear at the same position as the <1$\bar{1}$0>  of CaF2, as observed previously in MoSe2 on CaF2Vishwanath et al. (2015); Koma, Saiki, and Sato (1990). The lattice spacing of MoTe2 based upon the ratio of the RHEED streak spacing is $\sim$3.5 Å, which is very close to the value of 3.52 ÅKeum et al. (2015) corresponding to bulk 2H-MoTe2. The ambiguity in this measurement is due to the diffused RHEED pattern, which is measured more accurately using grazing incidence X-ray diffraction (GI-XRD), presented further below.

II.2.2 Raman Spectroscopy

Raman spectra in Fig.2 confirm an evolution from the 2H phase to a new phase as we progress from sample A to sample C. The 2H phase is confirmed by comparing Raman from sample A with Raman from CVT grown bulk 2H-MoTe2 obtained from 2D Semiconductors Inc. as shown in Table.2. Its important to note that the FWHM of the peaks from MBE grown MoTe2 is several times wider than that of the CVT-grown MoTe2. This points to a significant disorder in the MBE-grown material and augments the observation of the diffuse MoTe2 RHEED pattern in sample A. The new phase is assigned to the 1T’ phase due to its proximity to the peak position reported for 1T’ in literaturePark et al. (2015). So far, only 2H and 1T’ or Td phases of MoTe2 are known. As seen in Table.2 the Ag peak position of the 1T’ phase at 255.59 cm-1-256.23 cm-1 agrees closely with the reported value of 256.1 cm-1 - 257 cm-1 for 1T’ MoTe2 growth by tellurization of molybdenum films but the Bg peak at 158.07 cm-1-159.14 cm-1 deviates significantly from the reported value of 163 cm-1 - 161 cm-1 and extremely broadPark et al. (2015)Zhou et al. (2015) . Sample B shows a mixed phase comprised of Raman signatures from both phases, where as sample C only shows 1T’ peaks.

II.2.3 X-ray photoemission spectroscopy (XPS)

[FIGURE:]

XPS spectra corresponding to Mo, Te, O, Ca, F and C are detected from all samples (Fig.3). No charging effects were detected on any of them. Peak positions for Mo 3d5/2 and Te 3d5/2 as well as Te:Mo ratio corresponding to both phases 2H-MoTe2 and 1T’-MoTe2 are listed in Table.3. For sample A, the Mo 3d5/2 signal corresponding to 2H-MoTe2 bond was detected at 227.9 eV, which is consistent with the binding energy of 2H-MoTe2 in literatureJaegermann and Schmeisser (1986). The Te 3d5/2 peak corresponding to the 2H-MoTe2 is observed at 572.8 eV. Molybdenum oxide in the Mo+6 state was also identified. In the Te 3d spectrum, tellurium dioxide and MoTe2 are both detected. The Te:Mo ratio is $\sim$2.57 after correction due to attenuation from the oxide overlayer. For sample B, in addition to the peaks corresponding to 2H-MoTe2, Mo+6 oxide and TeO2 are detected; the Mo 3d5/2 peak at 227.6 eV and the Te 3d5/2 peak at 572.6 eV are assigned to the 1T’ phase. The Te:Mo ratio for the 2H-MoTe2 component is calculated to be 2.03. After correction due to attenuation from the oxide overlayer the Te:Mo ratio is calculated to be 2.54. The Te:Mo ratio for the 1T’-MoTe2 component in sample B is 2.12 and, when corrected for the oxide overlayer is 2.66. For sample C, there is a very small signal from 2H-MoTe2 with Mo 3d5/2 at 228 eV and the corresponding peak for Te 3d5/2 at 572.7 eV. But the majority of the MoTe2 peak intensity is from a new Mo 3d5/2 peak at 227.5 eV and Te 3d5/2 at 572.5 eV, which are assigned to chemical states associated with the 1T’ phase of MoTe2. The Te:Mo ratio corresponding to 1T’ phase of MoTe2 is 2.13 and, when corrected for the oxide overlayer is 2.72. The Te:Mo ratio for the 2H-MoTe2 component is 2.04 and, when corrected for the oxide overlayer, is 2.61. It is key to note that in sample C the Mo 3d peak intensity associated with Mo oxide is much higher than that for MoTe2, as well as the oxide intensity from the other samples. This suggests that, inspite of employing a large over pressure of uncracked Te (dimers) during growth, not only does Mo form predominantly 1T’-MoTe2 but that majority of Mo has an increased propensity for oxidation. This molybdenum oxide in sample C exhibits 2 different Mo oxidation states of +5 and +6. Reported peak position for 1T’-MoTe2 for Mo 3d5/2 is 227.7-228 eV and for Te 3d5/2 is 572.1-572.6 eVNaylor et al. (2016); Zhou et al. (2015). The observed XPS peak from the phase assigned to 1T’-MoTe2 for Te 3d5/2 is consistent with the reported value but that for Mo 3d5/2 is much lower that what has been reported for any Mo-Te bond and even metallic Mo 3d5/2 at 227.8 eVWerfel and Minni (1983). It is also noted that in all 3 samples, the oxide peaks from Mo and Te in the O1s spectra could not be resolved because of the close proximity in electronegativity of Mo and TeBernede et al. (2003). The O 1s spectral feature also has contributions from C-O and O-H. Fig.3(b) shows that concentration of both molybdenum oxide and tellurium oxide concentrations are higher on the surface as seen from the increase in intensity of oxides at take-off angle of 45$\degree$ as compared to 80$\degree$. (A take-off angle of 80$\degree$ is much more bulk sensitive than 45$\degree$.) Also, the chemical bonding state of MoTe2 is homogeneous through the analyzed depth, as the Mo-Te peak width remains constant when changing angle. Its worthy to note that the extent of oxidation in the telluride system is much more than previously reported MBE MoSe2Vishwanath et al. (2015).

II.2.4 Grazing incidence X-ray diffraction (GI-XRD) and Transmission electron microscopy (TEM)

In order to get a better estimation of the inplane lattice constant as compared to the estimation using RHEED pattern, and to understand the preference of in-plane rotational orientation, grazing incidence X-ray diffraction (GI-XRD) was done. GI-XRD from sample A shows (see Fig.4(a)) an extended line corresponding to overlapped $\{$10$\bar{1}$0$\}$ and $\{$10$\bar{1}$1$\}$ set of planes of MoTe2 and the sharp high intensity peak is from the CaF2 substrate. The in-plane lattice constant of 2H-MoTe2 calculated from the $\{$10$\bar{1}$0$\}$ peak corresponds to 3.638 Å. The full width half maximum (FWHM) of 2H-MoTe2 $\{$10$\bar{1}$0$\}$ peak is calculated to be $\sim$0.079 Å-1 and the direct beam FWHM is $\sim$0.009 Å-1. By subtracting the width of the direct beam, the genuine FWHM of the $\{$10$\bar{1}$0$\}$ peak was estimated to be $\sim$0.07 Å-1 which translates to a grain size of $\sim$92 ÅShimada et al. (1996). From Fig.4(b), which is an inplane phi($\phi$) scan, we can see that MBE MoTe2 undergoes significant twinning thus showing 2 sets of 6-fold symmetry diffraction patterns. The peak ratio between adjacent peaks separated by 30$\degree$ is $\sim$0.5. This shows that almost 30% of the grains are twins. Also the wide FWHM (6.7$\degree$-7.2$\degree$) of these peaks signifies a large deviation of grains from the preferred orientation. From measurements and simulation it has been shown in 2H-MoSe2 that $\{$10$\bar{1}$1$\}$ and $\{$10$\bar{1}$2$\}$ peaks are $\sim$10 times weaker than $\{$10$\bar{1}$3$\}$James and Lavik (1963). 2H-MoTe2 having the same crystal structure as 2H-MoSe2, we also observe the $\{$10$\bar{1}$3$\}$ set of peaks at higher q$\perp$ (see Fig.4(c)), from which the out of plane lattice constant (c-spacing) is calculated to be 14.4 Å. The c axis lattice constant obtained from cs-TEM as shown in Fig.4(d) is 13.9 Å. The reported value for the inplane lattice constant and c axis lattice constant from bulk 2H-MoTe2 are 3.52 Å and 13.966 Å, respectively Keum et al. (2015). To understand this discrepancy, we compare the intensity along q$\perp$ (see SI Fig.4) with simulations and observation for various polytypes of NbSe2 by Toshihiro Shimada et al.Shimada et al. (1994) 2H-MoTe2 and 2H-NbSe2 share identical in-plane crystal structures and hence, would give similar intensity profiles along q$\perp$ for various stacking orders (polytypes). Our q$\perp$ (see SI Fig.3) scan closely matches the one reported by Toshihiro Shimada et al.Shimada et al. (1994) on Se-GaAs, which is explained as a combination of 2Hb and 3R NbSe2. 2Hb and 3R both have the trigonal prismatic monolayer but the stacking sequence is different, with 3R having a 3-layer periodicity as compared to 2-layer for 2Hb. The value for MBE 2H-MoTe2 obtained by TEM is spatially local but X-ray beam for GI-XRD has a foot print of $\sim$2 mmx10 mm. The excess Te measured using XPS and the presence of stacking faults resulting in mixture of 2Hb and 3R phases could be the reason for the observed larger a and c lattice constants in MBE 2H-MoTe2 compared to bulk 2H-MoTe2.

SI Fig.3(a) shows the cs-TEM of the mixed phase sample B. For 1T’- MoTe2 is known to crystallize in P21/m space group with lattice constants of a=6.33 Å, b=3.48 Å and c=13.82 ÅKeum et al. (2015), where b is very close to the lattice constant of 2H-MoTe2. Therefore, for the sample C since the RHEED shows polycrystalline rings, irrespective of the in-plane rotational orientation of the film, we would expect to observe a ring corresponding to the $\{$010$\}$ set of planes in a similar scan as for Sample A (Fig.4(a)). SI Fig.3(c) shows the GI-XRD on sample C ie. 1T’-MoTe2. SI Fig.3(c) shows that there is no signal observed corresponding to 1T’-MoTe2, the only peak is corresponding to CaF2. This is likely due to extremely low signal from the polycrystalline 1T’-MoTe2 thin film. It also shows the variability in crystallinity of CaF2 from substrate to substrate, motivating the use of epi-ready GaAs substrates in Section B.

II.2.5 Electrical measurements

Sheet resistivity of the sample A and sample C were measured to be 5468 $\Omega$/$\square$ and 13255 $\Omega$/$\square$, respectively. This is of interest because film on sample A is 2H-MoTe2 and film on sample C is assigned to 1T’-MoTe2. Traditionally 1T’-MoTe2 should be metallic and have lower resistance than 2H-MoTe2 but here the extensive oxidation of Sample C could be the cause of the significant increase in resistivity. TeO2 glasses show semiconducting behaviorHampton et al. (1987).

III Section B : MoTe2 on GaAs (111)B

In this section, 2 samples (sample D and sample E) of MoTe2 on GaAs (111)B are discussed. post-growth anneal was done on MoTe2 films on GaAs.

III.1 Growth conditions for sample D

Sample D was grown at a growth temperature of 340$\degree$C and Te flux of 6.9x10-6 Torr (slightly higher than Sample A due to higher thermal conductivity of GaAs than CaF2). SI Fig.5(a), shows the growth sequence. The key step in this growth is the anneal of GaAs under Te prior to MoTe2 growth to achieve smoother Te terminated surface. This is consistent with our previous report of Te anneal of GaAsVishwanath et al. (2016) prior to TMD superlattice growth. The change in growth rate between samples (see Table.1) is due to variability of Mo flux at the same e-beam power.

III.2 Results and Discussion

III.2.1 RHEED

For sample D, a pair of faint RHEED streaks with a spacing less than MoTe2 were observed. Te has a hexagonal crystal structure with lattice constants of a=4.456 Å and c=5.921 ÅAdenis, Langer, and Lindqvist (1989). Since the inplane lattice constant of Te is greater than that of 2H-MoTe2, it was the initial suspect. With the aim to remove any excess Te in the film, a post-growth anneal at 380$\degree$C without any Te flux was done. But, as seen in SI Fig.5(b), the anneal doesn’t remove this second set of streaks. Further analysis reveals that the ratio of spacing of the RHEED streaks from MoTe2 and the newly observed streaks is $\sim$2. If the lines were from MoTe2 and Te, the expected ratio is $\sim$1.2. So, its likely not due to elemental tellurium at the surface. One hypothesis is presence of ordered defects, which could be Te interstitials. Fig.5(a) shows RHEED intensity oscillations of the RHEED spectral point during the growth of MBE 2H-MoTe2. This shows close to layer by layer growth. Simulation in SI Fig.7 shows the crest and the trough do not necessarily indicate a complete monolayer, variation in smoothness can cause shifts. But approximately, the period between crests corresponds to a monolayer. Increasing roughness or waviness in the film is the likely cause for decay with RHEED oscillation intensity in Fig.5(a).

III.2.2 Surface morphology, XRD and TEM

It is very interesting to note that after cooling to room temperature, sample D shows Te crystallites on the surface. From AFM image (Fig.5(c)) and SEM image (SI Fig.6(a)) we can observe that these crystallites have preferential crystallographic orientation with the underlying GaAs (111) with triangular symmetry. Also, we find that the height of these crystallites is $\sim$9 nm and about $\sim$50 nm wide. The fact that these crystallites are purely tellurium is confirmed by the markedly distinct lattice spacing compared to MoTe2 as seen in the high resolution transmission electron microscopy (HRTEM) image shown in Fig.5(d). This is further confirmed by TEM Energy-dispersive X-ray spectroscopy (not shown). The HRTEM image (Fig.5(d)) also shows a high quality of MoTe2 with a c-axis lattice spacing of 13.9 Å consistent with 2H-MoTe2. These crystallites could have likely been formed during the cooling process to room temperature. We don’t observe Te crystallite formation on films on CaF2 (see SI Fig.2). XRD scan in SI Fig.6((b) shows that at room temperature the (004) peak for sample D appears at 24.46$\degree$, which is lower than that for bulk 2H-MoTe2 and 1T’-MoTe2Keum et al. (2015). It corresponds to a c-spacing of 14.52 $\pm$ 0.05Å. It is analyzed below along with (004) peak of sample E (see Table.4).

III.2.3 Raman and electrical characterization

Since, the Raman signal from MoTe2 on GaAs is quite weakVishwanath et al. (2016), the as grown film was exfoliated and transferred to SiO2/Si substrate using a scotch tape. The transfer was performed to enhance the raman signal from the MoTe2 due to cavity effect from SiO2 as well as to eliminate the interference from the LO phonon raman peaks from GaAs. In Fig.5(b), the peaks below 150 cm-1 can be attributed to TePine and Dresselhaus (1971). E2g peak from transferred MoTe2 from sample D is almost symmetric and peak position is consistent with that from bulk MoTe2 at 235.8 cm-1. The reason for broadening in the A1g peak is unclear. In order to test the electrical characteristic of the transferred MBE grown MoTe2 film, contacts were made to the transferred flake. Using backgating, no modulation was observed (see Fig.5(e and f)).

III.3 Growth conditions for sample E

Sample E was grown on n$+$ GaAs for characterizations requiring conducting substrate. Prior to the post-growth anneal, the growth sequence for sample E was identical to sample D. The post-growth anneal for sample E was done at 450$\degree$C for 3 mins and 550$\degree$C for 7 mins under Tellurium flux. After this post-growth anneal, the sample was cooled under Te till growth temperature of 340$\degree$C and then capped with $\sim$100 nm Se during cool-down to room temperature for surface protection during sample transfer to other characterization tools.

III.4 Results and Discussion

III.4.1 LEED and XRD data

From the phase diagramBrewer and Lamoreaux (1990), one might expect mixed phase formation during the post-growth anneal at 550$\degree$C under Te. LEED and XRD was done to check the phase of the film grown. Prior to LEED, the Se cap was desorbed by annealing the sample in UHV at 300$\degree$C for $\sim$30 minutes. A LEED pattern measured with 40 eV electrons (Fig.6(b) shows two sets of spots with the outer hexagonal pattern corresponding to the lattice constant of 2H-MoTe2 (‘a’ lattice constant using LEED = 3.57 $\pm$ 0.03 Å)and the inner pattern corresponding to an effective 2x2 superstructure. This is very interesting because it is consistent with the second set of RHEED streaks seen for sample D in SI Fig.5(b), discussed previously. One possible explanation for the extra spots is a change in the surface periodicity relative to the bulk 2H-MoTe2 crystal structure due to a reconstruction or the presence of ordered defects. An alternate possibility is that the large electron spot size ($\sim$1 mm) may be averaging over three domains of 1T’-MoTe2 rotated by 60 degrees and 120 degrees (See SI Fig.8).

To distinguish between these possibilities, temperature dependent XRD was done on sample E. At room temperature (002), (004), (006) and (008) peak 2$\theta$ positions of the as grown film are 12.17$\degree$, 25.02$\degree$, 38.02$\degree$ and 51.38$\degree$ respectively. The reported room temperature 2$\theta$ positions for the (002), (004), (006) and (008) peaks for 2H-MoTe2 are 12.66$\degree$, 25.48$\degree$, 38.63$\degree$ and 52.34$\degree$ respectively. D.H Keum et al.Keum et al. (2015) report using temperature dependent XRD that the (004) 2$\theta$ peak of 2H-MoTe2 is at 25.5$\degree$ at room temperature and it shifts to slightly greater than 26$\degree$ at temperatures above 600$\degree$C corresponding to (004) 2$\theta$ peak of 1T’-MoTe2. In our case, the (004) 2$\theta$ peak position is much lower than both peaks ($\sim$0.5$\degree$ lower than 2H-MoTe2), which corresponds to a c-spacing of 14.25 $\pm$ 0.04Å, and the peaks from the film is lost above 400$\degree$C. Therefore, without a chalcogen over pressure, MBE grown MoTe2 dissociates between 400$\degree$C and 500$\degree$C. Also, sample D (see SI Fig.6(b)) which was grown without anneal at 550$\degree$C anneal has $\sim$0.5$\degree$ lower than sample E. Peak at smaller 2$\theta$ implies larger lattice constant but its origin is unclear yet. One possible explanation is the presence of excess tellurium in the crystal, which has been previously reported for bulk crystalsJames and Lavik (1963). The MoTe2 phase diagramBrewer and Lamoreaux (1990) shows that 2H-MoTe2 is not a line compound. The 2H phase of MoTe2 can be formed inspite of a 1% sub- or super- stoichiometric incorporation of tellurium. Hence, as XRD for both samples on GaAs does not show detectable peaks from the 1T’-MoTe2 phase, the 2x2 superstructure observed in LEED, is likely a surface feature rather than the presence of rotated domains of 1T’-MoTe2. Lattice constants obtained from the various techniques described above have been tabulated in Table.4.

III.4.2 Sensitivity to air exposure of MoTe2

To understand the ease of oxidation of MoTe2, on sample E, the Se cap was removed from Sample E by heating in a UHV system, followed by in situ XPS. After the initial XPS measurements, the sample was exposed to air for 20 mins and then XPS done again. It can be clearly seen from Fig.6(a) that oxide peaks appear in both Mo and Te XPS spectrum. Approximately, 8$\%$ of the surface area under goes oxidation in 20 mins. Therefore, through the process of exfoliation and fabrication the film is likely to undergo extensive oxidation. For TeO2 glasses the conductivity goes down with lowering temperature but the conduction mechanism is by charge hoppingHampton et al. (1987). The Se decapping is done in an oxide MBE system that is connected to LEED and XPS system to avoid air exposure.The O 1s signal observed in XPS (see Fig.6(a)) prior is air exposure is likely due to physisorbed oxygen, post-decapping, from the oxide MBE chamber (base pressure of $\sim$1x10-8 Torr). This is consistent with the fact that electronegativity difference between Mo and Te is 0.3 eV but that between O and Te is 1.4 eV, making the compound prone to oxidationJaegermann and Schmeisser (1986); Bernede et al. (2003). Effect of air exposure in MBE grown films could be exacerbated by oxidation at the edges of MoTe2 grains, similar to that reported for MBE grown WSe2 filmsPark et al. (2016).

IV Conclusion

This work employs extensive large area structural and chemical characterization of MBE grown few layer MoTe2, with complementing electrical characterization. In section A, we show that for growth of few layer 2H-MoTe2 at a low temperature of 340$\degree$C and growth rate of $\sim$6 mins/ML, we need an incident Te:Mo flux greater than 100. The 2H and 1T’ phases formed in different samples have distinct signatures in RHEED, Raman and XPS, but the Te:Mo stoichiometry determined by XPS is greater than 2 for both. GI-XRD shows a small grain size of $\sim$90 Å, twinning and a higher-than-expected c-spacing in 2H-MoTe2 on CaF2. XRD on MBE 2H-MoTe2 on GaAs in section B also shows larger c spacing than both bulk 2H-MoTe2 and bulk 1T’-MoTe2. Growth on GaAs shows Te crystallite formation on the surface, and a 2x2 pattern in RHEED and LEED. All these have been hypothesized as signs of excess Te incorporation into 2H-MoTe2 during growth. At ambient pressure in N2 atmosphere, MBE 2H-MoTe2 on GaAs is only stable up to 300$\degree$C. Excess Te in the film can explain the high electrical conductivity, non-modulating behaviour and easy dissociation of the film with increasing temperature prior to phase transition to the 1T’ phase, a more stable phase at higher temperatures. Finally, we demonstrate the swift oxidation ($\sim$8$\%$ surface area in 20 mins) of the MBE MoTe2 film on exposure to air. With the various complementing large area and local characterizations, this study has provided insight into the few layer MBE growth of Mo-Te system on 3D substrates.

V Experimental Methods

V.0.1 Raman Spectroscopy

Raman measurements were performed in the backscattering configuration using a WITec Alpha 300 system at room temperature. Measurement was done using a 100x objective, 1800 grooves/mm grating, 488 nm laser and 0.75 mW power.

V.0.2 X-ray Photoelectron Spectroscopy

XPS on the CaF2 (Series A) samples was carried out ex-situ using a monochromated Al K$\alpha$ source (h$\nu$ = 1486.7 eV) and an Omicron Argus detector (MCD-128) operating with a pass energy of 15 eV. XPS spectra were acquired at a pass energy of 15 eV and take-off angle (defined with respect to the sample surface) of 45$\degree$ and 80$\degree$. For XPS peak deconvolution, the spectral analysis software Aanalyzer was employed, where Voigt line shapes and an active Shirley background were used for peak fittingHerrera-Gomez, Hegedus, and Meissner (2002).

XPS on the 2H-MoTe2/n$+$GaAs (Series B) samples was measured using a non-monochromated Al K$\alpha$ source and a Scienta R4000 electron analyzer operating at a pass energy of 100 eV. All spectra were measured at normal emission, i.e. 90 degrees relative to the sample surface. Central peak locations were determined by Lorentzian fits with linear backgrounds.

V.0.3 X-ray Diffraction

Out of plane XRD and temperature dependent XRD on the 2H-MoTe2/GaAs samples is done using the Rigaku SmartLab X-Ray Diffractometer with Cu K$\alpha$ X-ray source. The GI-XRD is done using the G2 hutch at the CHESS beamline (http://www.chess.cornell.edu/gline/G2.htm), operating with a x-ray energy of 11.31 KeV.

V.0.4 Transmission Electron Microscopy

The atomic structure analysis for sample A and sample D was carried out on FEI Titan 80-300 Transmission Electron Microscope operated at 300 kV.

TEM on sample B was done using JOEL ARM200F atomic resolution analytical microscope.

V.0.5 LEED

LEED on the 2H-MoTe2/n$+$GaAs samples was measured using a Specs ErLEED 3000 system with an incident electron energy of 40 eV. The electron spot size was approximately 1 mm in diameter, and the total angular field of view was 100 degrees. Following the Se decapping at 300$\degree$C, in situ LEED and XPS measurements were both performed at room temperature in an analysis chamber with pressure below $\sim$1x10-10 Torr.

VI Acknowledgments

This work was supported in part by the Center for Low Energy Systems Technology (LEAST), one of six centers of STARnet, a Semiconductor Research Corporation program sponsored by MARCO and DARPA. This work made use of the Cornell Center for Materials Research Shared Facilities which are supported through the NSF MRSEC program (DMR-1120296). This work is based upon research conducted at the Cornell High Energy Synchrotron Source (CHESS) which is supported by the National Science Foundation and the National Institutes of Health/National Institute of General Medical Sciences under NSF award DMR-1332208. MBE growth is in part supported by NSF Grant DMR 1400432 and NSF-EFRI 2DARE Grant DMR 1433490.

VII Supplementary Information

VII.0.1 Possible origins of extra spots in low energy electron diffraction (LEED) patterns from bulk MoTe2

Fig. 8a shows the crystal structure of a monolayer transition metal dichalcogenide with undistorted octahedral transition metal coordination. Lines of metal atoms run in three different directions (referred to as 1, 2, and 3; see inset of Fig. 8b), where each line is separated by a distance d. In the distorted structure, shown in Fig. 8b, the metal atoms shift, creating zigzag chains of metal atoms where two neighboring lines are separated by a smaller distance d-$\delta$. These chains run along a single direction (in this case, direction 1), causing the unit cell to change from hexagonal to rectangular symmetry. In reciprocal space, extra Bragg reflections appear corresponding to the change in symmetry (Fig. 8c and 8d). Zigzag chains along directions 1, 2, and 3 are degenerate in energy, so it is reasonable to expect domains of different orientation in a macroscopic material. In this case, if the domain size is smaller than the electron beam, a LEED experiment will measure spatially averaged patterns from all three directions (Fig. 8e). The resulting Bragg reflections are identical to those caused by a 2x2 superstructure of the original undistorted crystal. Bulk 1T’-MoTe2 is formed by stacking individual monolayers with distorted octahedral Mo coordination. At a given incident electron energy, the out-of-plane periodicity determines the intensity of LEED spots but not the position of Bragg reflections. So a bulk 1T’-MoTe2 sample with multiple domains would produce Bragg reflections as shown in Fig. 8e. Alternatively, a bulk 2H-MoTe2 sample would produce Bragg reflections as shown in Fig. 8c. (The 2H structure has trigonal Mo coordination, but retains the same hexagonal symmetry assumed for Fig. 8c.) If the sample surface acquired a larger 2x2 periodicity through a surface reconstruction or the ordering of surface defects, the surface sensitivity of LEED would ensure that the resulting Bragg reflections would appear as in Fig. 8e. Because 2H-MoTe2 and the distorted octahedral structures exhibit the same in-plane lattice constant, the Bragg reflections measured by LEED would appear at identical locations to bulk 1T’-MoTe2 with multiple domains.

VIII References